Import experimental NS2 support and accumulated GPU fixes

Bring the NS2 patch and local follow-up fixes onto Eglantine master.
Includes native graphics services, instruction translation, syncpoint
cancellation acknowledgement, texture handling and diagnostics.

Experimental checkpoint; not ready to merge. The original Eden build
passed, but this rebased branch has not been built or tested.

Assisted-by: OpenAI Codex (implementation, debugging, tests and PR preparation)

1 files changed, 8 insertions, 1 deletions

diff --git a/src/video_core/query_cache/query_cache.h b/src/video_core/query_cache/query_cache.h
index b8c1bc4f45..c2fefa6695 100644
--- a/src/video_core/query_cache/query_cache.h
+++ b/src/video_core/query_cache/query_cache.h
@@ -20,6 +20,7 @@
 #include "common/settings.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/gpu.h"
+#include "video_core/native_query_trace.h"
 #include "video_core/host1x/gpu_device_memory_manager.h"
 #include "video_core/memory_manager.h"
 #include "video_core/query_cache/bank_base.h"
@@ -246,6 +247,9 @@ void QueryCacheBase<Traits>::CounterReport(GPUVAddr addr, QueryType counter_type
         return;
     }
     DAddr cpu_addr = *cpu_addr_opt;
+    const bool native_trace = gpu_memory->UsesNativeTiling();
+    const u64 trace_asid = gpu_memory->GetID();
+    if (native_trace) TraceNativeQuery(0, trace_asid, addr, cpu_addr, static_cast<u32>(counter_type), payload, static_cast<u32>(flags));
     const size_t new_query_id = streamer->WriteCounter(cpu_addr, has_timestamp, payload, subreport);
     auto* query = streamer->GetQuery(new_query_id);
     if (is_fence) {
@@ -262,14 +266,16 @@ void QueryCacheBase<Traits>::CounterReport(GPUVAddr addr, QueryType counter_type
     u8* pointer_timestamp = impl->device_memory.template GetPointer<u8>(cpu_addr + 8);
     bool is_synced = (Settings::IsGPUFenceBehaviorDefault() ? !Settings::IsGPULevelHigh() : !Settings::IsGPUFenceBehaviorBalanced() && !Settings::IsGPUFenceBehaviorAccurate()) && is_fence;
     std::function<void()> operation([this, is_synced, streamer, query_base = query, query_location,
-                                     pointer, pointer_timestamp] {
+                                     pointer, pointer_timestamp, native_trace, trace_asid, addr, cpu_addr, counter_type] {
         if (True(query_base->flags & QueryFlagBits::IsInvalidated)) {
+            if (native_trace) TraceNativeQuery(2, trace_asid, addr, cpu_addr, static_cast<u32>(counter_type), query_base->value, static_cast<u32>(query_base->flags));
             if (!is_synced) [[likely]] {
                 impl->pending_unregister.push_back(query_location);
             }
             return;
         }
         if (False(query_base->flags & QueryFlagBits::IsFinalValueSynced)) [[unlikely]] {
+            if (native_trace) TraceNativeQuery(3, trace_asid, addr, cpu_addr, static_cast<u32>(counter_type), query_base->value, static_cast<u32>(query_base->flags));
             LOG_ERROR(HW_GPU,
                       "Query report value not synchronized. Consider increasing GPU accuracy.");
             if (!is_synced) [[likely]] {
@@ -287,6 +293,7 @@ void QueryCacheBase<Traits>::CounterReport(GPUVAddr addr, QueryType counter_type
             u32 value = static_cast<u32>(query_base->value);
             std::memcpy(pointer, &value, sizeof(value));
         }
+        if (native_trace) TraceNativeQuery(1, trace_asid, addr, cpu_addr, static_cast<u32>(counter_type), query_base->value, static_cast<u32>(query_base->flags));
         if (!is_synced) [[likely]] {
             impl->pending_unregister.push_back(query_location);
         }